(EXT) RE: [PATCH 2/2] thermal/drivers/thermal_helpers: Adjust output format

[Alexander Stein]

Hello,

Am Mittwoch, 11. Mai 2022, 00:48:00 CEST schrieb Nitin Garg:
> On 08/04/2022 13:09, Alexander Stein wrote:
> >> Outputs like this where -1 is printed as unsigned is somewhat misleading
> >> 
> >>   thermal thermal_zone1: Trip3[type=0,temp=48000]:trend=4,throttle=1
> >>   thermal cooling_device3: cur_state=1
> >>   thermal cooling_device3: old_target=-1, target=2
> >>   thermal cooling_device3: zone1->target=1
> >>   thermal cooling_device3: zone1->target=2
> >>   thermal cooling_device3: zone1->target=18446744073709551615
> >>   thermal cooling_device3: set to state 2
> >> 
> >> With THERMAL_NO_TARGET assigning -1 as unsigned it make sense to print
> >> the target as signed integer, even if the type is actually unsigned.
> >> 
> >> Signed-off-by: Alexander Stein <alexander.stein at ew.tq-group.com>
> >> ---
> >> An alternative would be to change thermal_instance::target from unsigned
> >> long to long, but this would entail a lot of API & driver changes as well
> >> which looks less intriguing.
> >> 
> >>   drivers/thermal/thermal_helpers.c | 2 +-
> >>   1 file changed, 1 insertion(+), 1 deletion(-)
> >> 
> >> diff --git a/drivers/thermal/thermal_helpers.c
> >> b/drivers/thermal/thermal_helpers.c index 3edd047e144f..0d0da6670267
> >> 100644
> >> --- a/drivers/thermal/thermal_helpers.c
> >> +++ b/drivers/thermal/thermal_helpers.c
> >> @@ -199,7 +199,7 @@ void __thermal_cdev_update(struct
> >> thermal_cooling_device *cdev)>> 
> >>   	/* Make sure cdev enters the deepest cooling state */
> >>   	list_for_each_entry(instance, &cdev->thermal_instances, cdev_node) 
{
> >> 
> >> -		dev_dbg(&cdev->device, "zone%d->target=%lu\n",
> >> +		dev_dbg(&cdev->device, "zone%d->target=%ld\n",
> >> 
> >>   			instance->tz->id, instance->target);
> >>   		
> >>   		if (instance->target == THERMAL_NO_TARGET)
> >>   		
> >>   			continue;
> >
> >Actually you pointed out something fuzzy in the target values.
> >
> >The unsigned long type for the target and THERMAL_NO_TARGET are not
> >compatible.
> >
> >It would be much simpler to have THERMAL_NO_TARGET = 0 which
> >semantically makes more sense than a negative value.

Is it identical? Apparently target value is used differently in each governor. 
At least for gov_bang_bang 'THERMAL_NO_TARGET = 0' is no difference. Im not so 
sure about gov_step_wise.

> The compare of unsigned long and negative int is bad idea.

Well, THERMAL_NO_TARGET actually is an unsigned long (-1UL), so the comparison 
is unsigned long to unsigned long, so it should not be an issue.
But this implies that printing the target as unsigned int, results in a huge 
number, not immediately recognizable as -1, which I tried to address here.

> But there is serious problem introduced by "thermal: core: Add notifications
> call in the framework" patch. When system resumes from mem suspend first
> time (this happen only on 1st resume), the thermal notification is sent to
> drivers with value of 0 (meaning system is no longer hot). This is due to
> the fact target is init to 0 and when there is only 1 cooling device; it
> gets out of the loop (due to continue;) with target still set to 0 and
> calls thermal_cdev_set_cur_state(cdev, target). From there
> thermal_notify_cdev_state_update is called with argument of 0 which
> notifies drivers with value of 0.
> 
> May be "unsigned long target" should be initialized to THERMAL_NO_TARGET
> instead of 0.
> 
> [   29.107048] OOM killer enabled.
> [   29.110225] Restarting tasks ... done.
> [   29.124816] thermal cooling_device0: zone0->target=18446744073709551615
> [   29.138388] GPU0: Hot alarm is canceled.
> [   29.145399] thermal cooling_device0: set to state 0
> [   29.198954] PM: suspend exit

Is it legal to pass THERMAL_NO_TARGET to .set_cur_state()? At least pwm-fan 
will return -EINVAL in this case.

Alexander